6/13/2025 - Mastering ChatGPT Prompts: A Deep Dive into Project Workflow

Transcript

1. 0:00What do you think, boy? What do you think? [Music]
2. 0:23You and I all alone. Sunday morning here at home. [Music]
3. 0:40You and I here alone Sunday morning here at home. Blue and the coffee strong. It's true.
4. 0:55[Music] Then I open my eyes to a dream realized in front of me. And I haven't got a clue what in the world is happening to me.
5. 1:13Think I think I'm happy like first day summer vacation. Happy. Got to get a little rest and relaxation. Happy
6. 1:25like the choir on Sunday morning singing true. [Music]
7. 1:43Friday night date night at the AI learning lab. Are we going to learn tonight? No. We going to have a date tonight? Yes,
8. 1:52we're going to have fun. You never know. You never know. At the learning lab, sometimes there's learning. Sometimes there's comedy and
9. 2:02entertainment. A lot of times there's a dog. And then we've got Champy, too. Get it? I was calling myself a dog.
10. 2:17Um, that was comedy explained. See, we're we're when we're when we're teaching here, we're not just teaching
11. 2:27AI. We're teaching comedic structure. We're pointing out the obvious. We're making dad jokes about dad jokes.
12. 2:43All right. Fantastic. If it weren't for for Friday night date night, I wouldn't have a date at all. Oh, Kelly, you're working too hard. You
13. 2:54got too much AI in your life. We got to get you got to get you some of the people. Hey, Sharon Crawford, what's happening?
14. 3:02Oh, Kevin Stewart, does he always make jokes like this? Exactly. Yeah. Hey, you know about this channel that you're
15. 3:10watching? I No, I went No, I did. I I watched it last night. I did was this it was the AI learning lab with the
16. 3:21rotund the rounder gentleman the older round gentleman. Yeah. Yeah. No, I did. I saw it. It was um Yeah, it was um it was interesting. It
17. 3:33was interesting. Yeah. Mhm. And you watch that every night? Yeah. Yeah. No, I think No, I think for you that's
18. 3:40great. I think it's great for you. Yeah. No, it's it's it's I just you know what it is. I probably don't like AI. That's
19. 3:48probably what it is. I just Good evening everybody. All right. [Music]
20. 4:15[Music] Champy went out on a walk today. It was very traumatizing apparently because Champy got in a fight the other day
21. 4:29because he's a bully, but he got his ass kicked by by a uh not a go. What was it? A golden retriever? What was it that
22. 4:38kicked your ass? It was some dog that looked pretty, but he he looked prettier when he had his jaws around your neck.
23. 4:46Right, you big dumb dumb. Why'd you go pick on that dog? Huh? Now he's all scared today. Dogs are funny.
24. 5:01Oh, Champy, what are we going to do with you? Oh, Vicki. And no dirty nachos for me today. Oh, you had Sichuan and snooze. That's funny.
25. 5:16I love that for you. It was one of my least favorite phrases. Oh, I love that for you. Oh, you like that? Yeah. No, I
26. 5:22love that for you. That's fantastic. Yeah. No, it's for you. For you that's awesome. That's awesome. It's just, you
27. 5:29know, I just I have a different aesthetic. Maybe it's just you know I when I think of education I think of
28. 5:36more professorial kind of experiences if you with a curriculum and such you know with you know actual learnings. Yeah
29. 5:45that because the AI learning lab was um not that I mean it was it was an alternative approach to education which
30. 5:54I think is innovative. It was super innovative. Super innovative. Yeah. No, it was No, it was really I think you have a life.
31. 6:04Yeah. And it's And you're choosing to do things in it and I just for you I think that's really special. Yeah. No, I No,
32. 6:11it's it's all good. It's all good. I wasn't offended. There was a lot of cussing there. There was there was a lot
33. 6:16of cussing. Yeah. No, but it I But you know, I've got thick skin. I've I've certainly got thick skin for for you
34. 6:24know, people who choose to be that way. No, I I you know, they choose to live their life and and everyone should have
35. 6:31a space that you know I don't necessarily need to live in. You know, you know how that is, right? Yeah. Yeah. No, no, I do I do think
36. 6:40we're good friends. Yep. No, I do. I Yeah, I I I I know more about you now than I did than I did yesterday. So,
37. 6:47that's There's always a positive. There's always a positive. All right. I don't know why this turned into a an
38. 6:58improv of of one of the characters from Office Space, but you know, thank you for sharing the live Danielle
39. 7:07on a Tik Tok and whoever else has done that. I see some other people have done that. Thank you very much, Silver Fox,
40. 7:13as always. Kyle, what's up? What's happening? Kie, what is shaking? When someone asks, "Does he have any qualifications?" Yeah, about that.
41. 7:28[Music]
42. 7:53Girl, you're looking [Music] find a night. Every fellas got you in his sight.
43. 8:08What you doing with a clown like me? Wait. Surely one of life's little mysteries.
44. 8:24So tonight [Music] stars above how did I ever win your love?
45. 8:40Uh, wait. I don't remember the chords. That's a pretty song that I can't quite sing. It's too high for me. It's very high.
46. 8:52It's very high. Very, very high. Um, so a bless him. Exactly. That's exactly what it is, Vicki. It's good for Isn't
47. 9:03that sweet for you? It's totally bless him. A bless him. Ain't that ain't that sweet? you know, that AI learning lab,
48. 9:09you know, bless him. Bless him. He's out there night after night, um, you know, attempting to express himself. Just
49. 9:17bless him, right? You know, to each their own. To each their very own. Oh, yes. No, it was it was an experience. It
50. 9:26was absolutely an experience. Yes. No. No. It's Yeah. I Yeah. No, it was it was a Have I tried chat GPT? No. No, no, not
51. 9:38yet. Not yet. Not yet. Not yet. I You know, and you know, we just got rid of the fax machine, so slowly progressing. You
52. 9:48know, you know what I Yeah. No, still have the landline. Yeah, we still have a real phone. Okay. Yep. No, we still
53. 9:55have, you know, because I'll tell you what, when the power goes out, we can still make phone calls. Okay. All right.
54. 10:01That's why we have that. Oh, yes. Yes. Yes. Yes. Yesas. Now we know how to invoke the other Kyle. [Music]
55. 10:18[Applause] [Music] Um, happy Friday night everyone. So, uh, so I I kind of have a plan tonight, which don't get me wrong, it's my
56. 10:32version of a plan, so there's no actual plan. There's no actual curriculum, but so two nights ago, maybe two or three
57. 10:41nights ago, we made the Bigfoot video. Someone came in and asked about how do you make the Bigfoot video? And then I
58. 10:47said, you know, I started writing a prompt using my big fat brain. And then I was like, oh, wait a minute. Don't use
59. 10:54your brain. Have Chat GPT help you write that. And it did. And it came up with something really good. So I thought it
60. 11:00would be cool today. So on office hours today and then on the AI salon mastermind call which is right after it um
61. 11:12we had a lot of people talking and sharing really kind of deep deep uses of how they're doing prompting. Um and like Sam is it Sam?
62. 11:22What's Sam's last name? Sam it's not Sam Elliott is it? Is it Sam Elliott? That's an actor. But do we also have a Sam
63. 11:28Elliott? What's Sam's last name? Whatever. It doesn't matter. Sam was talking about this 16page prompt that
64. 11:33he's working on for work and I was like I like the concept of a one-page prompt intimidates me and he's got this 16page
65. 11:41prompt. So he was giving us all these best practices about how you do that. And then I think it was Vicki or someone
66. 11:49else was talking about um doing prompt generation within a project because you can keep adding documents and chats to
67. 11:57the project that contribute to it. So, I thought what might be cool is to go into chat GPT, have it come up with prompt frameworks,
68. 12:12put those in a canvas, export those as PDFs. So basically have it give itself instructions for how to write prompts
69. 12:20for things like scenes, characters, voice designs, things that we can, you know, start to stitch together components to make videos
70. 12:31and then put that all in a project and then we'll have a place where we can go start writing prompts and then we can
71. 12:36pull up things like 11 Labs voice designer or we can um pull up like Hedra or Runways character animation things
72. 12:45and like runway you can combine a location, some characters and oh and like a a a storyboard like a storyboard layout. So I just thought it
73. 12:58would be cool to make some sort of component maker. So we can try that. Now, what I know is that whenever I go
74. 13:05in with some sort of intention like we're going to do something that's a that's a real thing, it gets really
75. 13:11boring really fast cuz AI still sucks. But we'll try that. Okay. It is indeed Sam Samuel Elliott. Beautiful.
76. 13:26He is a a a very intelligent, driven, passionate young man. I like Sam a lot. [Music] Um Gareth, who's in a regular who comes
77. 13:42in here a lot, he said today that he's um he's gotten a new gig where he's like an AI he he basically
78. 13:53takes has taken his product development background and because he knows AI, he was given this sort of innov innovative
79. 14:02product development role that's about using AI to improve this company's products. So, further evidence of get
80. 14:10your ass curious about AI and and sort of bolted on to what you your core expertise and I think if you put that
81. 14:18combination in the marketplace, it's you're a different kind of hire right now, which is really cool. Kevin Stewart, I fly out to Japan to
82. 14:28Japan for a conference for AI next week. Oh, that's cool. That's awesome. That's really cool. All right, Francia.
83. 14:37Have you found chat GPT's answers over time are reflecting your own own inputs and biases? Yes, Kie. So,
84. 14:45if you've noticed that in the past six weeks or so, that's because about six weeks ago, maybe two months ago, Open AAI added um memory. Um,
85. 14:58so chat GPT now can remember I think it's your last year of chats. It knows them all. And if you're chatting, it will often
86. 15:11invade your current chat with stuff you've talked about before, which for a channel, for an account like mine is
87. 15:17really hard because I show so many different things on this channel that, you know, my my chat is very polluted
88. 15:24with with idiocy on my part. Um, so if you want to get rid of that, you can do it. Here, let me show you how to
89. 15:34do that. Hang on a sec. Let me get get things right here. I am This is the last night without producer Brandon and I think I have to
90. 15:43turn on the air conditioner because it's getting hot in here and I might pass out. Okay, so here's Chat GPT.
91. 15:50All right. In the upper right hand corner, go to your account and then go to settings and then in here, choose personalization
92. 16:01and you can turn off memory. There's two there's two kinds of memory. There's saved memories, which this was
93. 16:10the old memory. The old memory was you could explicitly say, "Hey, chat GPT remember this thing." And it would write
94. 16:16it to this file. And then sometimes it would just it it would kind of decide to to save a memory for you. So sometimes
95. 16:22you would write a prompt and it would say memory saved. You can go in and you can edit that one. This one below it,
96. 16:30reference chat history. This is the new one. You can turn that off discreetly. Um so if you want chat GPT to not be
97. 16:40influenced and biased by your previous chats, that's how you turn that stuff off. All right. So there you go. Look, the actual
98. 16:47education here, people, you see what I'm saying? It's not just a waste of time, this channel. It's mostly a waste of time. But it's
99. 16:55not all It's not all a waste of time. Just mostly. [Laughter] Oh, lordy, lordy, lordy, lordy, lordy.
100. 17:11Do you know any songs? I Yeah, I know. I know. I know the the uh the 12 songs. I know the 12 songs of Christmas. Um, I
101. 17:22know lots of songs in my iPad, but the ones I actually remember are [Music] [Applause] [Music]
102. 17:38this place I can rest my paid to gather my thoughts in sweet silence. And this is this place where the
103. 17:53feelings on dead [Music] from an overexposure to violence. This is a place I can slowly face. The only one I truly can know.
104. 18:11These are tears from a long time ago. Got these tears from a long time ago. I need to cry. That is a song. These are tears for a long time.
105. 18:28Go. Um, caveat. I only know depress depressing songs. That one's called 30 Years of Tears by
106. 18:43John Hyatt. incredibly prolific, talented songwriter, performer, amazing amazing talent. Uh most of his songs are
107. 18:53just they'll just drive you to drink. Oh man, when do you start taking requests? When I learn more than 12 songs.
108. 19:07What education? I'm out. I know, Kelly. That's not why we're here, is it? Okay. Um, the other thing I thought I I had a very
109. 19:20busy day today. Busy, busy, busy. I was a busy boy. So, normally I try to find an hour or two during the day to surf
110. 19:29Twitter and see if anything launched that is is worth looking at. And I didn't do that. So, it might be fun to
111. 19:35just just go scroll Twitter for a bit and see if there's anything interesting there. Um, M I'm here strictly for the
112. 19:44entertainment which that that's saying a lot on on many levels. I took a pill in a beza. Nice. Joy party's in the house. Joy party in the
113. 19:57house. We could make images. We can make images. Well, again like what I want to do I I am my I am newly I'm newly committed
114. 20:12[Music] I'm starting to get my head around. I've been saying this for a while. 2025 is the year I'm going to stop using my brain. [Music]
115. 20:26[Laughter] And I want to put some context around that. [Music]
116. 20:43So, okay. So, let me give you my theory on on on creative energy. Let me turn on the air conditioner first. Hang on.
117. 20:59All right. So, one of the things I've learned over my many, many years, and this is true for me, this might not be true for
118. 21:07everyone. I've got ADHD. I'm neurospicy. So, that might have something to do with it. But my experience with creativity
119. 21:18is that on any given day, there's a fixed amount of creative focus that I can that I can engage with.
120. 21:32And I I I don't know dayto day how long that's going to be. Some days I wake up and it's six or seven or eight hours
121. 21:39where I can just focus and do like a whole day of [ __ ] Some days it's like 20 minutes. And what I know is that and
122. 21:48and when I when I think about creative mindset or create creative engagement, it doesn't mean I'm making pretty
123. 22:03pictures. For me, creativity is problem solving. And so, it's like, okay, I got this thing I want to do. And so, it might be
124. 22:11I want to make a picture, I want to make a video, or I want to write something or I want to do this. And then it's like
125. 22:15it's all these minations about, okay, how am I going to do that? What am I going to do? What do I got to do?
126. 22:22And then on any given day, like the amount of energy you have to do that is, for me, it's fairly fixed. And if you
127. 22:29spend, let's let's say I woke up today and I've got three hours in the tank. If I spend those three hours problemsolving
128. 22:40on something that's a low-level problem, like how do I get this this this word processor to give me something formatted in a way
129. 22:51that I don't have to format each line? Let me learn stylesheets and then I go learn stylesheets. If I spend three
130. 22:57hours problemolving on low-level tactical [ __ ] then I have no energy left for the higher level truly creative thinking.
131. 23:11And and what's struck me over the past two years working with generative AI is that it's really good at some of the low-level
132. 23:26blocking and tackling of creativity. It's really good at it. And so very often someone will come in here and say,
133. 23:33"Hey, how did well that this is what happened the other night. How are people making those Bigfoot videos I'm seeing
134. 23:39all over X?" Which I didn't know what they were. We went and looked at some. We said, "Ah, okay. I know how to do
135. 23:44that." And then I started to sit down and say, "Okay, I want a video about a Bigfoot." And then I was thinking about,
136. 23:50"Well, where do Bigfoots live? They live in the Northwest and they Right. And so we probably want something and we
137. 23:56probably we we want him looking at a selfie. So like, how do you describe selfie?" Like I I started going down the
138. 24:03rabbit hole of doing all the problemolving of figuring out what elements should be in that prompt. And then what hit me was this moment of
139. 24:15no, back out. Zoom back out. Don't get seduced into using your brain for low-level [ __ ] because all of that stuff isn't really
140. 24:28creative. It's kind of tactical, right? It's like, okay, what kind of camera shot should it be? And what kind of
141. 24:36So, I zoomed out and then I said, "Hey, chat GPT, there's these videos with Bigfoots doing selfies. Can you write me
142. 24:43a a prompt for for a a a video tool that does that?" And it wrote this. It wrote what's on screen right here,
143. 24:54which is not super long, but it's it's cool structure. It said, "Here's your scene setup. Bigfoot is holding a shaky selfie style,
144. 25:03a a shaky phone selfie style. He's midbrant, sweaty, slightly out of breath with a squirrel on his shoulder that he
145. 25:11ignores completely. Like like that that line with a squirrel on his shoulder that he ignores completely.
146. 25:20That's just good writing. That's just good writing. Right. In fact, when we made this video, someone in the comments said, um, "Is
147. 25:30that a squirrel on his shoulder? Um, you get one continuous take. No cuts, no effects, just raw Bigfoot
148. 25:40energy like that. That's just there's a lot going on in here creatively that I can look at and I can appreciate and and
149. 25:48I could do one of two things. I could look at it and go, that's exactly what I want." Or I could look at that and go,
150. 25:53"Oh, no. It completely missed the mark." But see, now I'm just I'm operating at a higher level. I'm operating at like the producer level.
151. 26:02And so the the thing I want to start to get my head around and I want to put into this this new um project I have called Feed the Prompt is
152. 26:14the ability to let AI free you from the low-level crap that chews up all your energy. So you can just you can live at this higher plane
153. 26:25of creativity and just just be like dictating down into the details. No, I want it more like this. I want it more
154. 26:32like that. A little sprinkle of this. A little sprinkle of that. Um and then I went in and gave an acting
155. 26:39prompt. Look dead into the camera. Whyed and unblinking with frantic intensity. Deliver this line. I eat organic. I
156. 26:49compost my poop. And you all still think I'm a myth. and it capitalized that. This is a funny line, right? I didn't have to come up with
157. 26:58that archetypal architect. That's pretty much how I go through go about writing. I see and have a lot of images and
158. 27:07videos. They're cool, but I'm now stuck on how to use these to make a difference in my life with my work. Using them for
159. 27:15work sucks. It doesn't work. Interesting. You know, Valerie, um, Cindy [ __ ] sent me a I don't know if
160. 27:23Cindy's on the call tonight. It's Friday night date night. She might be. Um, Cindy sent me a video that she made
161. 27:29today in an hour in um in LTX Studio and she's making things for work and she's making things for clients and
162. 27:42she's doing them incredibly quickly. So, so I think here's the trick, Valerie, is one is you got to play with the tools
163. 27:53enough to understand what's possible. And I think you do that. You're here a fair amount. But if you haven't played
164. 27:57with LTX Studio, there might be something there. There might be also something about connecting with Cindy
165. 28:01just to see see how she's doing it. I actually suggested that Cindy do a learn out loud on on this one project that she
166. 28:08showed me today. um because it's got a lot of janky crap in it, but it doesn't matter because it's it's just basically
167. 28:15this this I don't know. It it was it was just this cool thing that communicated an idea in video and it did it it did it
168. 28:23quite well and it didn't have to be perfect. Um but I think I think the trick especially when you're getting into video stuff, the the
169. 28:34problem with video is video is hard. Video is complicated. It's structured. It's layered. You've got to be good at,
170. 28:43you know, visual stuff and audio stuff and musical stuff and sound effects stuff and acting stuff and writing
171. 28:50stuff, right? And directing stuff and and compositing it together stuff. There's there's a lot of stuff you have
172. 28:57to do. So, it's very easy to to run down rabbit holes with with video where you're letting the tool lead rather than the
173. 29:08idea lead. And I think I think the trick with video is you got to be really clear on what you're trying to accomplish for
174. 29:13what audience and then it gets a bit easier. But but it's um it's like that you're having trouble with it at work doesn't surprise me
175. 29:24necessarily because if that's not what you do regularly for work, it's going to be this foreign territory. And I but but
176. 29:32again I think the key is start with what are you trying to accomplish? Who's the audience? Um I'll reach out. Yeah.
177. 29:38Great. Yeah, I would definitely connect with her. Um, okay. Archetypal Architect, you definitely watch that, Lol. Yeah,
178. 29:47because I think Cindy, we we've done a couple of, you know, things with LTX Studio where we've gone in and said it
179. 29:56can do this and it can do that and it can do that. But what I was thinking about when Cindy sent me this one
180. 30:02specific output is I want her to do a lull that starts by sh showing the output and then go all the way back to
181. 30:11the beginning. What was she trying to accomplish? Who was the audience? Who was the client? What was she trying to
182. 30:18accomplish? And then what was her thought process? Did she start in chat GPT? Did she go to Claude? Did she go to
183. 30:25Perplexity? Did she go straight into LTX Studio? When she got in there, what did she do? How did she pick a style? Right?
184. 30:32Like, if you want to knock something out quick in LTX Studio, that flumxes me because it's so complicated. There's so
185. 30:39many switches and dials. I get lost in all of that. She's figured out a way to navigate through that really quickly.
186. 30:46So, I think that would be quite valuable. Um, this would help me greatly in how I think about video, I think. Yeah, like like video again,
187. 30:57if you start with I need a video or I need a video that's like this style and then try to work your way back to the
188. 31:06story, it's really hard. You really have to start with, you know, what am I trying to accomplish? What's the story?
189. 31:12And then work forward to what's the right tool because there's just too many tools. Um, What did Leonardo put out today?
190. 31:28I use 4.1 for therapy and recipes. I love that. The New York Times and Gizmodo today. Brandon sent this to me. Um had a thing
191. 31:40about how how AI is ruining people's lives. I wish VO3 had a prompting guide. Yeah. Well, so V3 probably does have a prompting guide,
192. 31:56but it's probably um it's probably like the history of cinema is the prompting guide, right? Like, so in fact, let's
193. 32:10go. Let's let's Okay, let me go I'm going to pull out here for a second. So, I created a project called the video scene builder.
194. 32:20And so I've got the Bigfoot thing we did the other night is in here in here. But let's start a new chat in here. And let's say um research
195. 32:33VO3 from Google and tell me everything you know about it. um and recommend the best way ways to
196. 32:51prompt it to give results that are professional and consistent across
197. 33:08scenes. And then I'm going to turn on Let's [ __ ] it. Let's do a deep research. We'll we'll flip over to 03.
198. 33:25So So I'm in the model 03, which is one of the reasoning engines. If If you're How many folks are in here? 34. We
199. 33:32probably have some folks in here that are new. If you're new to chat GPT, GPT40 is a less sophisticated model than GPT 03.
200. 33:49Okay, you're like, but why? And then 04 mini is better than 03, but it's mini, so it's not as good as 03. So
201. 34:0203 is the big model that does reasoning. 40 is the sort of daily driver. So we're gonna we're gonna have it go do some
202. 34:11deep research and really think on the best way. VO3 shock collar. Shut up. 03 Pro is going to blow minds. Yeah, I
203. 34:23don't have access to it yet. Okay. So let's let's have this go do some work. Could you please clarify what you mean by V3?
204. 34:33Are you referring to specific video generation model? Yes, it's VO3, the new video gen model used in flow.google.
205. 34:55All right. So, it's starting research. capabilities, features, architecture, best practices for prompting, generate
206. 35:06professional, consistent results. So, here's here's what I'm here's what I'm hoping we can do. And I think you can now share
207. 35:18projects. So, this could be a project that I I share on the on the salon. This is I'm gonna need a drink type of thing from chat GPT
208. 35:35starting research. Listen, I don't think we need to do deep research on this thing. Like I think we could have just
209. 35:41done a basic web search, but I always think it's good to, you know, use tools that are overkill for for the for the challenge at hand.
210. 35:51Um, how did you find or get that scene builder? I just made a project called scene builder. I that's that's I literally
211. 36:03just said like 10 minutes before I came on the live, I had an idea for what to do tonight. I thought so. One of the
212. 36:10things we talked about in office hours and in the mastermind um meeting today was was using projects as a way to to
213. 36:24augment pro prompting. So here's here's one of the things I learned today from from the the sort of group conversations
214. 36:30that we had today. Um, most people that are writing sophisticated prompts are having chat GPT or whatever the, you
215. 36:45know, their LLM of choice is helping them write the prompts, right? So, they're saying, "Hey, I need a structured prompt that accomplishes
216. 36:54this or here's a big long prompt." Sam Sam Elliott was talking about he got hired into this job and he inherited
217. 37:02this 16page really sloppy messy self-contradicting prompt and he put it into chat GPT and he said reorganize this remove anything
218. 37:14where it self-contradicts and put it in structured organized sections that make the most sense to accomplish XYZ.
219. 37:23And now he's maintaining this much more organized, you know, structured thing. But how he's getting all of the
220. 37:30components and improving things is using chat GPT. He said like if he wants it to write a better email like one of the
221. 37:39sections is write good emails he'll do a test where it says write an email and then if he doesn't like the output
222. 37:51he'll copy the output paste it back into chat GPT and say okay here's the prompt here's the output it gave me and the you
223. 37:59know the prompt I gave it the output it gave me and here's why I don't like it rewrite this section ction of the big
224. 38:06prompt so that it improves that email. And then he says it just will rewrite stuff for him and then he just keeps
225. 38:14doing iterations like that. So I I'm kind of I've never really prompted that deeply and I think it's a good skill to
226. 38:25understand and yeah it is this is technically prompt engineering but I think you can you can vibe prompt engineer
227. 38:39which is if you know vibe coding right you just ask for what you want and these coding tools just code for you. We can
228. 38:45do the same with with prompting sophisticated prompts. Do you think do you have the project published yet? Don't know how they work. Lol.
229. 38:57That's the first time. Yeah. Um I've heard of it. Oh, okay. So, so let me show you where projects are.
230. 39:05Projects are kind of cool. So, so you've got custom GPTs, right? And you can make one of those. projects are kind of like
231. 39:13a custom GPT, but they're kind of different. So, let me go in. So, here's one I've got for Sydney. So, Sydney's my musical. So, you can
232. 39:23upload a project file, right? So, I've uploaded the script. I've got custom instructions. Act script and story
233. 39:31consultant for Broadway musicals. Don't let us write crap. Everything has to be award-winning. I don't know if that's a
234. 39:37good system prompt, but that's what it is. And then here are all the different prompts, all the different chat
235. 39:43conversations I've had about Sydney. So now whenever I create a new chat inside this project, it takes all of this other
236. 39:51stuff into into consideration. The one that I just started for video scene builder really has nothing in it,
237. 40:00right? It's got it's got the Bigfoot comedy vlog prompt that we did and it's got this research that we're doing right now.
238. 40:10So that's it. And there's there's the research it's doing. But now what's cool is this whatever we end up doing in this chat, this chat
239. 40:20itself now becomes source information for the project. Make sense? So, think of a project. The project is
240. 40:30really just a folder with lots of chats within it. But because you can give it custom instructions and because you can
241. 40:37upload um data documents, source documents, it's kind of like um like a like a less specific version of Notebook LM. Notebook LM's like very
242. 40:53very specific to what it does. This is sort of like that. It's sort of like a custom GPT. Um but but it's pretty cool stuff.
243. 41:04And it it just got more powerful. Like they've added deep research. You can have a conversation within it. You can
244. 41:11share them. You can do all sorts of things. So, okay. I've been having the AI prompt itself since 3.0. Yeah, that's great. That's awesome.
245. 41:22I posted it in the salon. Leonardo put out a V3 prompting guide today. Oh, cool. Whi which channel did you put put it in, Vicki?
246. 41:33So, if you if you don't know what the salon is, the AI salon, go to go to that address, the salon.ai, and click on join our community. And
247. 41:47that'll take you over to the community site and introduce yourself. join, jump in, and then there's all sorts of different
248. 41:56spaces in there. Um, [Music]
249. 42:15so
250. 42:30Um, where did Vicki put it? Tool talk. All right. Tool talk. So, this is behind the mastermind. It's in the mastermind
251. 42:41area. Beautiful. Um, so within the salon, there's a free area of the salon and then there's a subscription area
252. 42:47called mastermind that has some some deeper level stuff in it. One of them is tool talk. So, Leonardo just published a
253. 42:53guide for prompting in VO3. Very cool. Mastering prompts for VO3. Let's go grab this. Planning your prompts. Okay, here's a
254. 43:07really good example of the the old way we used to do things with our brains and the new way we're going to do things with our with our
255. 43:19brains augmented by these tools. So, somebody wrote this VO3 prompt guide, right? Now, historically, what
256. 43:28you would do is you would print this thing out, right? And then you would have it beside you and you'd sit down to
257. 43:34write a prompt and then you go, "Okay, I got to write a scene description. Then I got to write a visual style. Then I got
258. 43:40to think about camera movement." And if you don't know camera move lingo, that would be a Google search to go
259. 43:47understand camera movement in Hollywood and cinematography. Then you've got to describe your main subject. And is the main sub subject a
260. 43:55character or a car or a building or a cloud? Ah, now like like at this point I'm starting to get overwhelmed and
261. 44:03break into a flop sweat. And then you got to get your background setting. And then you got to get lighting right. And
262. 44:08then there might be an audio cue and a color palette and dialogue and subtitles and language, additional planning notes,
263. 44:15writing your prompt. Are you [ __ ] kidding me? So that's the old way. The old way is we would get overwhelmed.
264. 44:23Oh, VO is on Leo. You can Google it if you're not in the mastermind. Yeah, Leonardo. Prompts. You know, VO VO is on Leo.
265. 44:33That's funny. Check it out. Um, but let's copy this. Go back to chat. Jeep. All right.
266. 44:49So, here's our deep research result. This is big. Look at this. I like it. Good lord. It's a lot.
267. 45:04Download PDF. Great. Perfect. Okay. So, it it it researched for five minutes. It found 19 sources and 34 searches.
268. 45:18This This is your brain. This is your brain on AI. If you know, you know. Actually, you know what, Kelly? I think
269. 45:24I'm going to steal that if you don't mind. I'm going to put that in my in my new presentation. [Music]
270. 45:39So, we're going to do two things here. Okay. Technical capabilities. Great. Supported inputs, text, images, and controls.
271. 45:56The primary input for V3 is a text prompt, but you can use images. In Google's flow interface, it can leverage ingredients,
272. 46:09dolly, zoom, aerial, wideshot, generation, quality. So, it's got some background [ __ ] Camera and scene control scene builder
273. 46:18for continuous scenes outpainting. Okay, great. So, we're going to download this as a PDF. All right. So, we now we now have the
274. 46:34deep research as a PDF. So, let's go back out to the top level of our project and we're going to upload we're going to add files. So, we're
275. 46:49going to add this thing that it just wrote. How many pages is this? One, two, three, four, five, six, seven,
276. 47:01eight. eight pages of of how to use V3 from deep research. We'll assume it's good. It's probably not, but let's just assume it's good.
277. 47:15All right. So, we have that in there now. Um, let's see. I'm going to say based on the knowledge
278. 47:34You have now and this article about the best way to prompt VO3,
279. 47:50there's the article. So, so I just pasted in the article that someone just g that Vicki gave us for VO3. Not
280. 47:58someone, Vicki. Vicki gave us it. I'm bad at that. People give me stuff and I'm like, a person somewhere in the
281. 48:07universe provided this URL. Thank you, Vicki. All right. Um, comma, I want you to
282. 48:24create a structured prompt recipe.
283. 48:44Feel free to use JSo N JSON if you feel that is important. A structured prompt recipe and guide
284. 49:07that captures everything a filmmaker
285. 49:21might need. to create both individual
286. 49:38clips but also consistent um shots throughout a story.
287. 49:54Okay. So, so to be clear, and who was I think it was I think it was Kelly Camp. Yeah, it was Kelly Camp today. This was in the
288. 50:07mastermind conversation. Kelly Camp was sharing she uses chat GPT to help her write prompts. But what she
289. 50:15does is she kind of vomits out all of the kind of requirements for what she wants in a prompt, which is that's
290. 50:23essentially what I'm doing here, right? I just uploaded some knowledge, so use that. Here's an article. Use that. And
291. 50:30now I want you to give me a recipe and I want you to give me a guide. And then, right, and so let's just let's just
292. 50:36generate this thing. And let's say we're going to say um why don't you put the resulting guide in a canvas? And if you don't know
293. 50:51what a canvas is in chat GPT, it's a separate editable document. So if if you use chat GPT and it gives
294. 51:00you like some big long thing and then you don't like one of the paragraphs within it for the most part, you have to
295. 51:07like rewrite the whole thing. If you if you put it in a canvas, you can just rewrite little pieces of it.
296. 51:15All right. And so I forgot I was in 03, but that's okay. We're going to do a reasoning model version of it.
297. 51:25Oh, look. There's our our single JSON template. Good [ __ ] lord. This is a lot. Like this is overwhelming. The anatomy of a V3 prompt.
298. 51:43V3 recipe and filmmaker guide. Scene description. Scene underscore description. So when I asked it to give
299. 51:50me a structured prompt, what it's doing is it's writing all of the fields that will be described in the JSON. So
300. 51:59JSON is just like a markup language. It's JavaScript object notation I think it stands for. And large language models are very very
301. 52:10good at structured language, right? Because a lot of what they teach these things to do is code. So it's good at
302. 52:17structured language. So, it's defined all of our different things. Color palette, dominant hues, LUT, which is a
303. 52:24lookup table. Um, deep slate, gray, ocean blue, and these are example snippets. All right. Enclose dialogue and straight
304. 52:33quotes if you want it spoken. Omit quotes if you don't want subtitles. Tip. Look, look at that tip right there. One
305. 52:40of the frustrations of um V3 is that it adds subtitles and they're never right. The words are always wrong, but it adds
306. 52:48them. If you put things in quotes, you'll get subtitles. So, this included this in our little thing here. So,
307. 52:56here's a single clip JSON template. Scene ID, intro pool scene, scene description, a glamorous woman with an
308. 53:03afro hairstyle and gold jewelry shares a nighttime pool with a calm alligator. All right, great visual life. Visual
309. 53:11style cinematic 35 millimeter film. So, this is great. Okay. Field notes. All right. Multi-seene continuity pack. Great.
310. 53:22Iteration and troubleshooting. Um, quick reference cheat sheet. Okay. So, here's something that's missing from here. when we did our um
311. 53:36our Bigfoot video clip the other night. It gave us acting notes. So, so let's see. Um
312. 53:52the other night I did a Bigfoot parody. Bigfoot selfie parody. And it's pair OD, not pair A D, pair O
313. 54:09D. And the prompt had
314. 54:23character and acting notes in it as well as the technical visual elements. Um, your guide
315. 54:40seems to ignore acting and character all together.
316. 54:56Recommend how we incorporate incorporate this into
317. 55:10our guide. So again, I could think through, well, you could create it as a you could create character as a separate module
318. 55:21and but [ __ ] it. Let it figure it out. It's the reasoning model. Go off and reason, chat GPT. Don't take up my
319. 55:28precious [ __ ] brain cycles. [Laughter] Kak Kyle, see my notes above about Bigfoot. Okay. [Music]
320. 55:44I'm laughing at these vids. They are hysterical. Here, I'll show you that one. Kak, there is now a medium foot and a yeti
321. 55:54complaining about the beavers. That's [ __ ] awesome. Yeah. Like like I gotta tell you, man, we're absolutely coming into this era of
322. 56:10joy where someone does something creative and other people can immediately respond to it, not with shitty comments, but with their own
323. 56:18version of it and then like Yeah. Like like Bigfoot and medium foot. That's [ __ ] awesome. Okay. Anatomy of the
324. 56:26big. So did it. So let's see. I expanded the guide to include performance focused fields, performance notes for emotion, gesture, comedic
325. 56:37beats, plus JSON keys, continuity packed, iteration, and troubleshooting. Chris reference cheat sheet. Um uh I think you also need to
326. 56:57include options for dialogue in the absence of dialogue. Um,
327. 57:13VO3 will provide it, but it would be nice to leverage powerful
328. 57:27LLMs to write better dialogue and have more control before
329. 57:42generating the video again. I'm going to say, "What do you recommend?" Let it do the [ __ ] thinking. I don't
330. 57:54need to use my brain to do your stupid work. What was the link for that research thing, Vicki? I'm only finding the
331. 58:05Leonardo article on the salon. I didn't I don't know which research thing you were talking about.
332. 58:20IP for no one. Yeah, exactly. Well, Joy, you know, it's funny, Joy. So, IP for no one, intellectual uh property rights for no one. Um I I
333. 58:31really do think we are entering a postcopyright world. Not that you don't want to protect your IP, but your ability to protect it becomes
334. 58:41increasingly diluted. Um, and it's already completely diluted, but it's going to get ridiculous, right? And I
335. 58:49think what happens is like right now the thing about copyright is that all of copyright law is about the output,
336. 59:02right? Because when they wrote copyright laws, it was actually difficult to copy things. So you'd produce a thing, an output
337. 59:13and then if someone copied it, that was the producing of it was really valuable and the the copies were hard and then
338. 59:23they got easier and easier and easier. Now we're moving into a world where there it's trivial like fan art is now
339. 59:31going to be as good or better than the original work. And so I think what has to happen is you protect the copyright on the generation,
340. 59:43not on the output. You you do it on the input, not the output. And that's where you do attribution and that's where you
341. 59:48do compensation. I think that's the way it's going to go. I think it's the way it's got to go. But who knows? Who
342. 59:54[ __ ] knows? Okay. Dialogue options are now fully woven into the guide. There's a dedicated dialogue script with
343. 1:00:02key sub fields. Okay. Um, let's see. Let's let's also add in also right now VO3 clips are only 8 seconds long. So
344. 1:00:18this guide should um guide the user to create prompts
345. 1:00:34that are not that are um possible within that constraint. Constraint All right, beautiful.
346. 1:00:53Oh, the deep research you saved as a PDF. I was in and out of the room. I may have missed a step. Oh, it's it's the
347. 1:01:01it's it's the one I saved. The the PDF I saved. I haven't shared that yet. I think what I'm going to do is I'm going to share
348. 1:01:09I'll just rewind. Yeah, don't mind me. I'm pretty high. That's okay. Listen, you're high. I'm inarticulate, but okay, great. Um,
349. 1:01:24there's a couple of things. I think we can share projects now. So, I'm going to share this project in tool talk in the
350. 1:01:31mastermind. So, anyone that's in mastermind will have access to this whole project. um that specific thing, that specific
351. 1:01:40PDF that I did based on deep research. I don't actually know if it's any good. I I can maybe I'll I'll pop here's what
352. 1:01:48I'll do. I'll put I'll put the project in mastermind tool talk tool tips. I'll put that that PDF that I generated in
353. 1:01:58irregulars because I I haven't really vetted it. Um I'm vetting this little thing I'm making a bit more.
354. 1:02:06All right. dialogue script. Yo, humans, just finished my morning jog. Feeling legendary. Okay, great. So, that's good. So, let's see. So,
355. 1:02:18runtime reality check. V3 will not generate clips longer than 8 seconds. Plan your storytelling and tight beats,
356. 1:02:26one action at a time. Okay. Two. Single clip JSON template. Same as before with seconds. Wait, but it it ate all of our
357. 1:02:36Oh, okay. Wait. In the canvas, you have now truncated the JSON outputs
358. 1:02:55as same as before, etc.
359. 1:03:14return. We need this canvas to include the entire
360. 1:03:28prompting guide. Please return the detail. You wait, you ninny, [Laughter]
361. 1:03:46for people who missed office hours, the summary is now on 100 flies on the wall. Oh, that's awesome, Vicki. That's really cool.
362. 1:03:53I'm excited to see it. All right, what's going on over here? I so for the if anyone's modding on Tik Tok, I know I'm
363. 1:04:00not paying attention right now and there's a lot of comments in here. Um, so I'm probably missing stuff.
364. 1:04:11Wait, I don't have the power anymore. Hope he sees it. If Is anyone modding? If is if anyone's modding, pin something that I should be
365. 1:04:20looking at. And if not, I'll just scroll back and we'll do this the oldfashioned way. Remember when I used to scroll
366. 1:04:26through all the comments and I would comment on them and people are like, "Is he ever going to get to mine?"
367. 1:04:33I don't think we have any anyone modding tonight. Modding every Oh, Vickiy's modding everywhere. Okay. All right. So, let's
368. 1:04:40see. The canvas now contains the full unabbridged guide. Every JSON field, continuity trick, dialogue workflow.
369. 1:04:48Okay. There's that. Leveraging LLMs for tight dialogue. Great. Okay.
370. 1:05:08All right. So, now let's download this as a PDF. And then we're going to close this. And we're going to go back out to our
371. 1:05:19project scene builder. Did I do this as Wait, where did
372. 1:05:33where did my that last conversation just go?
373. 1:05:48Did we just [ __ ] lose a whole chat? I think we did.
374. 1:06:06Huh? I can mod now. Okay. Kak new subscription model. Is there if there is content I used to see that I no longer have access to?
375. 1:06:25You mean in in the mastermind and the salon? Yeah. The the the clubs and um some of the mastery workshop things like
376. 1:06:36tool tips and things like that are now part of the mastermind that are behind the payw wall. So, yes, some some stuff
377. 1:06:43isn't there. And then there's new things that we're putting in there like libraries and things like that.
378. 1:06:52All right. All right. So, I'm going to add one more project here. Add files. So, now we've got this prompting guide.
379. 1:07:08Okay. Boom. Okay, two files. I don't know if I want custom instructions yet. I'm kind of pissed that I lost that last chat. Where did
380. 1:07:24that go? Let me reload here. Maybe something's funky with my browser. Video scene builder. Ah, here we go.
381. 1:07:37Yeah, it was a screen redraw thing. All right. So, here's my So, now I've still got that prompt in there. That's or that
382. 1:07:43that chat in there. Okay. So, we now have three different chats. We've got our our Bigfoot chat. We've got our
383. 1:07:53research that generated the the the VO thing. And then we've got the the prompting guide that incorporated that
384. 1:08:01PDF. And we created a prompting guide. And then the prompting guide itself became one of these files. So, we now
385. 1:08:08have two very focused files in here, right? We've got the the V3 deep research thing on prompting and we've got our own generated
386. 1:08:18prompting guide that includes all the video [ __ ] plus all the acting [ __ ] Okay. So now we should be able to say,
387. 1:08:34I want you to create a V3 prompt
388. 1:08:52for Um, the pilot scene, the pilot 8second scene
389. 1:09:08of a new parody of Bigfoot selfies.
390. 1:09:26called microfut selfies where
391. 1:09:40microfut beasts are evil looking but tiny
392. 1:09:58and have a short guy complex. [Laughter] Give me the first prompt
393. 1:10:12complete with funny dialogue to kick off the series. All right. So in theory,
394. 1:10:26in theory, because we've done a project and because we've got good high quality guidance in it, what this thing's going
395. 1:10:39to produce is a structured probably JSON prompt that has good comedy in it. Let's Let's see what it does. Here we go. This could be bad.
396. 1:10:52[Laughter] Oh, great. Google needs to hurry with the extend feature, but I can wait. The users creating to Okay, I've I
397. 1:11:08I should have turned off 03. So, by the way, 03 is a reasoning model, so it does thinking, so it's not as fast. So, if
398. 1:11:16you want to turn [ __ ] around quickly, um, you should probably just flip it back to 40. All right, there's our prompt. Why it
399. 1:11:27works? It uses every field. Okay. So, let's let's copy this and let's flip over to this.
400. 1:11:48An Microfoot. An evil 6-in crypted launches his first microfoot selfie vlog, glaring into his phone with world domination swagger.
401. 1:11:59Performance notes: Napoleon complex strut, exaggerated arm swings, ominous chuckle, halfbeat pulse before the
402. 1:12:07punchline. This is good dialogue, script, microfut. Hey, giants, I'm 6 in of pure doom. Kneel before me, redwoods.
403. 1:12:16Now hand me a ladder. This This lens insults my stature. Okay. Delivery. Raspy gremlin baritone. Cocky 110 per
404. 1:12:27minute sharp emphasis on doom and ladder. All right. Visual style handh held found footage selfie horror comedy vibe. This is I
405. 1:12:39mean this is Blair Witch, right? That just described Blair Witch Project. Handheld found footage selfie horror comedy vibe. That's cool.
406. 1:12:53Arm length selfie closeup. Subtle bob as creature hops up on a stump. Lens 14 millimeter fisheye. So you see you see why
407. 1:13:10my resistance to use my brain is paying off. This is a level of detail that I just wouldn't I I just don't give enough of a [ __ ]
408. 1:13:19Right. But if you want to get really good results, you have to like be a pro. Well, you don't have to be a pro. Let the AI be the pro.
409. 1:13:28I watched a movie called Exists today about Bigfoot. It's a found footage movie. There you go. So, we could just
410. 1:13:34actually create one of those. All right, let's go make let's go make this [ __ ] All right, we're gonna we're gonna see if this thing works.
411. 1:13:42Cam Ken, that's amazing. Isn't this cool? Right. So, so just to be clear, what we started with was a prompt from two days ago
412. 1:13:54where we had chat GPT take a basic idea and make it better. And then we said, "Hey, use your deep research and go find out
413. 1:14:05everything you can about V3 and write me a guide for for everything you know about it, including how to prompt it."
414. 1:14:12And then Vicki found this Leonardo thing. And then we said, "We want to we we want a guide to be able to do structured prompts," which we
415. 1:14:24then modified because we had some experience. And so now we've got this project that has all this intelligence and I'll just
416. 1:14:32be able to keep coming back in here anytime I want to do something for V3 and throw it in here. Let's let's go see
417. 1:14:37if this actually works. Can you explain the difference between 40 and 03? Yes. Yes. Okay. Wait. Let's go. You know what we'll do?
418. 1:14:52Let me go. Let me jump over to flow. Let's get this rendering and then I'll explain the difference between chat GBT 40 and chat GBT3.
419. 1:15:04It is actually [ __ ] criminal at this point that they're still allowing these naming conventions to to exist to
420. 1:15:14coexist. It's really bad. Um, but that's neither here nor there. We're going to do text to video. One of the important
421. 1:15:23things with with flow is that this always defaults to non um nonV3. It defaults to V2. So you got to flip it to V3.
422. 1:15:38And then oh and by the way there's V3 fast and there's V3 quality. All right. So V3 quality is 100 credits.
423. 1:15:51BO3 fast is 20 credits. So, I'm going to go with that. Now, I'm going to paste in. There's all my JSON.
424. 1:15:59You don't need to know what JSON is. You don't need to know how to write it. It's basically just tagged. It's tagged text
425. 1:16:08is all you need to know about it because it's our our little prompt machine is going to write it for us. All right. So,
426. 1:16:14we're going to get that going. Okay. So, while that's generating that, let me just change back to this mode. Okay.
427. 1:16:22Because I got to talk with my hands. Um, a little history.
428. 1:16:37So, when when chat GPT came out, it was Chat GPT 3 was what they called it. And then the model sitting underneath it was GPT3 and
429. 1:16:49then it was GPT3.5 and then I think at some point they called it chat GPT 3.5 and then they called it chat GPT4.
430. 1:17:06So that so the early progressions was the numbers were just sequential. It was 33 3.54 and then and and chat GPT4.
431. 1:17:19This this is probably a little bit too much background, but I I hopefully it will explain what's going on. Chat PT4
432. 1:17:31was what they call a mixture of expert models where it was actually 10 different models that they they sort of added together to create um
433. 1:17:46to create a single model. So when you would ask it a prompt it would say which of these like one of them was good at
434. 1:17:51coding, one of them was good at writing, one of them was good at whatever. There were 10 different expert models that
435. 1:17:56that it it sort of dynamically stitched together. The reason that's relevant is it was a 1.8 trillion parameter model. It was 10 whatever it
436. 1:18:08is 180 billion parameter models stitched together. It was very big and very expensive. And then they came out with chatpt 4 and
437. 1:18:20the O stood for omni. And omni means multimodal. So it could do images, it could see images, uh it could do the voice thing and it
438. 1:18:31was much smaller. So they were getting a similar quality of output with multimodal capability um
439. 1:18:47for for way cheaper and that became the default model. So for a long time there was GPT40 which was this tight omni model and then
440. 1:18:57sitting underneath it was GPT4 which was this big ass behemoth thing. They've since retire retired GPT4 I
441. 1:19:06think. Let me go just make sure it's not still there. Yeah, it's it's gone. So So four is gone. 40 was is is kind of the daily driver model.
442. 1:19:2003. This is just [ __ ] bonkers. They have 40 and then they said, "Hey, we figured out
443. 1:19:35a a new way for this large language model to behave." So they took sort of a base model and then rather it
444. 1:19:48giving you a single result for a single input they created this model called 01 not 10 O 01. So 40 the O stands for omni
445. 1:20:0401 the O stands for open AI1. It was the 01 reasoning model. And what the 01 model did is when you would give it a prompt, it would it
446. 1:20:18would prompt and answer itself multiple times before it actually answered you. It's what they call reasoning. It's not
447. 1:20:25actually reasoning from a definitional uh point of view, but but the all of the models that have O in front of them are
448. 1:20:36reasoning models. Okay. So, so we had 01 and then you had 01 mini and 01 mini high. So, 01 is the big model and the
449. 1:20:47mini models are smaller and not as expensive. They skipped O2 because someone else had the trademark. So, 03 is actually O2.
450. 1:21:02But it's the O in front of it that lets you know it's a reasoning model. And the 40 is an omni model. Here, let me show
451. 1:21:09you one other thing. Let me flip back and share share my screen. Um, so, so you've got 4 O. The O after the number means it's it's an it's a
452. 1:21:26multimodal model that can understand images, generate images, understand audio, generate audio, generate text,
453. 1:21:35generate code, all that sort of stuff, understand all that, do reasoning, write code, right? So it's it's a big ass
454. 1:21:41fancy model. 03 is the one that when you prompt it, it says thinking and then it says, "Okay, the user wants to do this.
455. 1:21:49Now I'm going to go do some research. Now I found this. Now I'm going to tell the user this. So it does all this back and forth stuff.
456. 1:21:5604 Mini and 04 Mini High are are the small versions of 03 or or of the next model beyond 03 that we're getting kind
457. 1:22:08of a a sneak peek into what's coming. And then down here you've got GPT 4.5 and GPT 4.1 and 4.1 mini.
458. 1:22:18None of these are are multimodal models, but they are sort of the next just different flavor. It doesn't [ __ ] matter.
459. 1:22:28The important thing to know is just these two. GPT40, that's your daily driver. 03 is the reasoning model. And
460. 1:22:35then if if you want to get more sophisticated, you can play with the other ones. But just know that those two
461. 1:22:39things are different. If you want a fast response, choose 40. If you want a a thinking response, choose 03. And 03, to
462. 1:22:47be clear, is a more sophisticated, not a more sophisticated model. It's a very different kind of interaction than you get with 40. So
463. 1:22:58sometimes 03 is really good for something. If you just want like a quick headline or something like that, 40 is way better.
464. 1:23:05If you wanted to do stuff like go research everything you can on VO and write me a report. That's why I went to
465. 1:23:11to 03 because it's using the reasoning. All right. Make sense? I know it's it's it's [ __ ] insanity. So is so so wait.
466. 1:23:21So is 04 mini this is great. My 40 has started thinking. So two block two block Tom the 40 starting to think there's
467. 1:23:30speculation that they're sneaking some of what's coming in GPT5 into 40 right now. So if your 40 says thinking they're
468. 1:23:40starting to put some reasoning into the the the base model. What they have promised us and I I I give this zero [ __ ] credibility at
469. 1:23:51this point. What they have promised us is that GPT5 will get rid of this nonsense that GPT5 will be a single model that's
470. 1:24:01got an orchestration layer that that it figures out which model it should use underneath Cam Katkin. So is 04 mini better than 03
471. 1:24:12or is it just faster? Okay, 04 mini high some people say is better than 03.
472. 1:24:26So, so it is definitely the next base reasoning model, but it's the mini version of it. So if you think 03 I I
473. 1:24:36don't know what their parameter numbers are at this point but let's just say let's just say 03 is you know 500
474. 1:24:43billion parameters 04 mini is probably you know a 100red billion parameters or or 50 billion parameters right it's it's
475. 1:24:52a way smaller model but it is the next it is the next model right so apparently 04 mini high is better at coding than 03
476. 1:25:033, but 03 is also really good at coding. So, I think this is one of those things where experiment with it. If if you're not
477. 1:25:14getting the result out of 03, try 04 mini. Try 04 mini high. Now, they also did just come out with 03 Pro, which is
478. 1:25:23only available on the pro account. And that's one where I think what makes it pro is they just run it longer.
479. 1:25:33So, so when you put a prompt in 03 Pro, it runs for like 15, 20, 30 minutes before it gives you a result. Like 03
480. 1:25:42will run for five minutes and then give you a result. 03 Pro is just running much longer. So, but apparently 03 Pro
481. 1:25:52is delivering really high quality results. So, good flipping Lord. It's, you know, I'd forgotten about All about
482. 1:26:0001. I know. Well, and now like now that you've got 40 and 03 03, you know, people think that three is worse than
483. 1:26:11four because that's how software's always [ __ ] worked. You you always go by the number. if they had called it like
484. 1:26:22GPT reasoning 03 or GPT reasoning 3 and then just you know um GPT for multimodal or something like that like
485. 1:26:34if if they just given them some name that says this is a different thing I don't think people would be confused like what the [ __ ]
486. 1:26:47can no one does does no one there have a liberal arts degree that can say this is [ __ ] stupid. I mean, it's staggering to me. They're a
487. 1:26:59[ __ ] what? $3 trillion company at this point. They can't hire a a branding consultant for a week. I'll do I'll
488. 1:27:09we'll [ __ ] do it right now. 03 Pro. Will 03 Pro run scenario situations? Probably. So like 03 03 Pro is when when
489. 1:27:25you want to solve big ass problems. So I would think so because one of the things that 03 has that is mindblowing
490. 1:27:39is it can use tools. And not only can it use tools, it writes its own tools. So if if in its thinking it realizes it
491. 1:27:49needs to do some statistical analysis on some data it found it will write and execute Python code to do the
492. 1:27:56statistical analysis as part of the reasoning. So I would think that 03 Pro does that on steroids, right? It's probably crazy.
493. 1:28:07Honestly, I think I prefer this to something like how they name car models or engineering model numbers.
494. 1:28:14you prefer what to it? What they've done because I think what they've done is how they're naming car numbers, right? It
495. 1:28:21does seem like they could have just followed the frigin software conventions or just giving them a [ __ ] name like like the like GPT
496. 1:28:29reasoning three. You've got GPT4 and GPT reasoning 3. That would solve it. Well, it wouldn't solve it, but it would at least be clear.
497. 1:28:44You know, you have GPT4 with pictures, GPT reasoning 3, GPT 4.5 creative writing, GPT 4.1 fancier but not quite
498. 1:28:57as cool as GPT4 pictures. Like, just, you know, give a dude a break. All right, let's go see our video.
499. 1:29:13Oh, that's cool. Remember we we said it was a like a What did it say? That it was a fisheye lens. 10 mm fisheye lens.
500. 1:29:34Oh [ __ ] We got um we got subtitles. A giants. I'm 6 in of pure doom. Kneel before me, Redwoods. Now hand me a
501. 1:29:44ladder. This lens insults my stature. A giants. I'm 6 in of pure doom. Kneel before me. Okay. So, let's go back to
502. 1:29:52our chat jpai and say um I'm gonna say listen, you had the rule that if you include
503. 1:30:09quote marks, it would generate subtitles. It did. And I don't want subtitles.
504. 1:30:23Also, the microfut is holding a phone.
505. 1:30:38we can see and it should look more like he is filming with a phone we can't see.
506. 1:30:54Update the prompt, you ninny. Oh [ __ ] I'm not sharing my screen. Shut up. Shut up. All
507. 1:31:14right. What did it say? It didn't say anything. No subtitles. No visible phone. Copy. Go back over to flow. Flu. Hello. Hello. And flu. Flu.
508. 1:31:28Bang. All right. So, that thing's generating. All right. We'll wait for that. Um,
509. 1:31:45use the selfie stick stick in the prompt. Yeah, that's pretty good idea. All right, beautiful. All right, let me
510. 1:31:54go back and look at Tik Tok comments because there's a lot of stuff going on in Tik Tok. Tik Tok's being really janky
511. 1:32:00for me. So, night all. Uh, all right. Bye, Silver Fox. Sorry, it's being janky. 4.5 is creative writing. Yeah. So, 4.5
512. 1:32:18is is not multimodal, so it doesn't do the image generation stuff. The rumor that I've heard is that 4.5 is actually going to be the primary
513. 1:32:33mediating layer of GPT5 that it's it's really good at empathy and understanding instructions and it's it's apparently quite good at creative
514. 1:32:43writing. So, if you're having not good results with 40, you should absolutely try 4.5 for creative writing. It's
515. 1:32:52probably worth experimenting with just to see what it does better. But yeah, it's a creative writing model.
516. 1:33:02We need an open AI model bingo card. We really do. It's so bad. Looking to get into a rabbit hole tonight with Zapier. That's fun. Can't
517. 1:33:14they use chat GPT to design a branding campaign? Seriously, I'm a librarian. They they could they could [ __ ] knock
518. 1:33:20this out in an hour. Like we could do it right now. See, this still gave us um
519. 1:33:52All right, we got to figure out a way to fix this dialogue thing. So, let's go look at our JSON. We'll find our dialogue dialogue
520. 1:34:04script. All right, let's get rid of the prompts. Hey, Giants. All right, let's just try that. Um,
521. 1:34:20go. We need an open AI model bingo card. All right, let's see what else we got in here. Um, if they only had a reasoning
522. 1:34:30model to help with naming conventions. Exactly, Sasha. It's like, seriously, what is going on? Ask Gemini. Yeah. Uh
523. 1:34:49you can get V3 for free so long as you as you have Google Workspace. You you don't get that many generations.
524. 1:34:57They don't have qualifications. Sahil, can I ask you a question? Yes. Ask away. Did you already ask a question? I'm
525. 1:35:13because I was focused on building us a thing. I didn't see your question if you asked it, but yeah, ask away. Horrible Park.
526. 1:35:34All right, let's see. Can we start a change.org petition to take over AI open AI naming? I don't think I I apparently with GPT5 we're not
527. 1:35:49going to need it because we we won't won't have to deal with the models anymore. But here's the problem. This is
528. 1:35:56a problem with all software. They're like, "Hey, listen. GPT's 5 is coming out in a month or two. So, let's we'll
529. 1:36:03just keep the names the way they are and and it it won't be that long because we we'll have GPT5 out by the end of April
530. 1:36:10or the beginning of May. Well, now we're in the middle of June and GPT5 is not coming out I think until July and that may push till August.
531. 1:36:20And then since then they've launched GPT 4.5, 4.1, 4.1 high, 04 mini, 04 mini high on top of 03.
532. 1:36:36For a while they they they had 03 mini, 03 mini high in there. And for a while they had 03 mini high with 01 and 01 pro. So, in the past
533. 1:36:49three, four months, we've had 10 models on top of the core model and the one below it, the 03 model. Won't be that
534. 1:36:59long. And yet, we'll have five more model names. I know, I know, I know, I know. This is why they need a cranky Gen
535. 1:37:06Xer in the corner office at OpenAI. Hey, Sam, I'm here. I know you're jealous of me. Just put me down the hall. Uh, I'll come scream at you
536. 1:37:17occasionally. Hey, Sam. Uh, [ __ ] knock it off with the names now. Well, we got 05. Yeah, but it's never gonna
537. 1:37:24launch. So, could we just [ __ ] fix it short term? Just rename them so that people aren't losing their [ __ ] minds
538. 1:37:30and not using the right models. I'll be cranky in the corner for free. I got my family on Grock because chat GPT
539. 1:37:43is so confusing and complicated with all the models. Exactly, Robbie. Exactly. Like there, you know, they got to get
540. 1:37:50their [ __ ] together. It's like needing to have a dozen open tabs. They won't be able to hire just one model. Well, in
541. 1:37:57theory, the the new model will do all the all the routing. And but you know what they found out? why chat GPT hasn't launched or GPT5
542. 1:38:08hasn't launched yet. They found out that even chat GPT can't figure out what [ __ ] models to use in what situation.
543. 1:38:18So like they're saying to us, here's a consumer app, chatgpt, right? You go to chat.com and it takes you to chat GPT. That's a
544. 1:38:28consumer play. The thing that made Chat GPT compelling when it came out was you didn't have to pick models. You didn't have to do the
545. 1:38:37temperature dial. You didn't have to understand what coding was and what a system prompt was. You just stuck in
546. 1:38:42words and out came an answer. And it was [ __ ] brilliant. They got to 100 million users in six weeks. And now
547. 1:38:50we've got eight [ __ ] models to choose from. They don't understand who their customer is. Like it's just it's [ __ ] insanity.
548. 1:38:59[ __ ] insanity. All right. Good lord. All right. So, it will be like a switchboard. It's literally going to be like a
549. 1:39:11switchboard, Gavin. It's It's going to be one ringy dingy, one ringy dingy. Yep. It's going to be Lily Tomlin at the
550. 1:39:18switchboard plugging in wires. So, when we type something into chat GBT, it's going to go like, "Hm, I wonder which
551. 1:39:25model would be best for this." I know. We'll use a combination of two models. Click. Click. I hope. I hope it does that. All right.
552. 1:39:38Hang on. I got to change my audio settings or you won't be able to hear. You'll be like, "Kyle, we can't hear the
553. 1:39:43little gremlin. We want to hear your microphone. We can't hear it. What's going on? This is not very fair to us
554. 1:39:50people on YouTube." All right, here we go. It got the phone in the frame, which I don't like, but whatever.
555. 1:39:56Giants. I'm 6 in of pure doom. Kneel before me, Redwoods she. Now hand me a ladder. This lens insults my stature.
556. 1:40:04Eight giants. I'm 6 in of pure doom. Kneel before me, Redwoods sheep. Now hand me a ladder. This lens insults my
557. 1:40:12stature. Giants. You know what? I want to try that same prompt with the higher quality model. Um, so I think I do that. Yep, there we
558. 1:40:25go. But then I'm going to do higher quality model. We'll we'll spend uh we'll spend 100 credits on this one. All right, there we go.
559. 1:40:40I forgot how to use a GPT on the chat. Wait. Oh, can you please show us how to use Canva GPT? Do you mean
560. 1:40:53do you mean you mean canvas within chat GPT cosmic? Did you already talk about lovable being
561. 1:41:08free this weekend? I didn't Rick Olsen. That's a good reminder. Thank you. So all weekend this weekend lovable.dev is
562. 1:41:15free. you can do. They might rate limit it if their servers start melting down, but they're doing a contest this weekend
563. 1:41:23and they just kicked off this thing called Lovable Shipped, which is a I think it's a six-w weekek accelerator
564. 1:41:30program for people that applied to it that said, I want to do cool [ __ ] with Lovable. So, that kicked off, but
565. 1:41:35they're kicking it off with it's free this weekend. So, if you haven't tried vibe coding, if you don't know what vibe
566. 1:41:41coding is, vibe coding is basically where you just speak a web app into existence or any app really depending on
567. 1:41:52the tool. Lovable just produces web apps. So, so web applications, but you can literally just say, "Make me
568. 1:42:00a clone of the game Asteroids," and it will make that or make me a recipe database that people can log in and put
569. 1:42:07in their family recipes and it will go build you the entire app and then you it builds it and then you're like, I don't
570. 1:42:14like this about it and it'll go fix that. So, it's free all weekend. If you've never tried it, it's [ __ ] remarkable.
571. 1:42:24If you've never tried vibe coding, it's absolutely insane. GPTs, any GPT like the Sin City Sipper. Okay, Cosmic Lover,
572. 1:42:31I'll show you how to do that. Um, let me let me let this uh well, no, let me let this video render and then I'll hop over
573. 1:42:39and uh and show you quickly how to how to create your own custom GPT. In two years, I'm going to vibe code a
574. 1:42:46smartphone operating system. Yeah. Well, I you know what's funny, Robbie? I think like by the time we get to two years, I
575. 1:42:54think software development is going to be an entirely personalized dynamic thing. I think software as we know it
576. 1:43:04will effectively cease to exist. It'll it'll sort of atrophy away. And what's going to happen is as we need software,
577. 1:43:14we will sort of birth it into existence. It'll do what we need it to do and then it'll go away. Um there may be
578. 1:43:21persistent things like data but I think even even things like databases like a CRM I could imagine that CRM dynamically
579. 1:43:30reconfigure themselves. There's a core data set and then how they behave is dependent on what you need in the
580. 1:43:38moment. I like I think we go to an incredibly personalized place. Source camp, thank you. I haven't really needed
581. 1:43:45Claude much lately. I'll give it a try. We might not have smartphones in two years. That's true. We might not. What's
582. 1:43:52free all weekend? Lovable.dev. Lovable.dev. Just need a lot more hardware and infinite energy. Yeah, exactly. So,
583. 1:44:10so, so Pate who's in the in the chat, you know, works for Google on on optimizing Tensor there. Oh, man. This added flipping.
584. 1:44:25Dang it. Dialogue. Let me generate another one. It added subtitles to it.
585. 1:44:41Eight giants. I'm 6 in of pure doom. Kneel before me, Redwoods G. Now hand me a ladder. This lens insults my stature.
586. 1:44:49Eight giants. I'm 6 in of pure doom. Kneel before me, Redwoods G. Now hand me a ladder. This lens insults my stature.
587. 1:44:57Eight giants. I'm 6 in. You know what's cool though with this script? Because it's got all that detail in it. Like the
588. 1:45:06acting here is really consistent because the acting notes are it it's hitting the right. It said hit the words ladder and
589. 1:45:13doom and it it did which is amazing. It must be in training.
590. 1:45:28What software are you using for that video? This is um this is Google's V3. Um, I'm at flow.google. flow.google is
591. 1:45:39so um, so Google partnered with Darren Arnowski, a filmmaker, and a bunch of independent filmmakers. So when they
592. 1:45:47created the V3 model, they they they brought in folks from Hollywood that, you know, didn't want to burn the building down
593. 1:45:55because some of some of the Hollywood people are pretty pissed right now. But some of them are like, oh, like James
594. 1:46:00Cameron's like, oh, here's a new tool for self-expression. What's most important is the story. It doesn't
595. 1:46:06[ __ ] matter what tool you use, right? So, so some of them get it. So, Darren Arnav Ornowski, I think is how you say
596. 1:46:14his name, um, is one of those filmmakers. And then they brought in a bunch of independent filmmakers. Flow.google is a really hyper
597. 1:46:24lightweight scene editor where I can generate these clips. And you'll notice up here in the lefthand corner of each
598. 1:46:31clip, it says add to a scene. I can stitch these things together and make a little movie. Um, it it doesn't really
599. 1:46:38have much in the way of sophisticated timeline editor stuff. It was pretty cool. Um, all right. So, let's see. That thing's
600. 1:46:47still going. Let me just show you quickly. Um, Cosmic, here's So, if I go to a new chat. So, here. Oh, wait. I got
601. 1:46:56to change my sharing. See, this is why I didn't want to flip out of here because everything's a [ __ ] production. All
602. 1:47:04right. Um, so here you are at chat GPT. On the left hand side, you've got, you know, your your basic elements at the
603. 1:47:13top. You have whatever you have access to like code, codeex, Sora, and then you've got your custom GPTs, right? So,
604. 1:47:20here's my custom GPTs including there's Sin City Sipper. Um and let's see
605. 1:47:34GPTs. Yeah. Okay. So, you click on GPTs and then that gives you the big long list. This is the the chat GPT store, right? So, you can find
606. 1:47:45other people's GPTs. You can search here but in the upper right hand corner you click on create GPT and that'll take you to this dual window
607. 1:47:56model where you can just say I want a GPT that you know um help helps me calculate when the next full moon is and it will
608. 1:48:06it will do something and create it here and then over here in the right you can play with it and then so you can either
609. 1:48:12do this in create mode where you're you're literally just talking to it like like a chat GPT chat and it'll build you
610. 1:48:20the custom GPT or you can go into configure mode where you upload a picture, give it a name, a description,
611. 1:48:27here's your prompt, um here are the different tools it has access to. You can upload data files to it, things
612. 1:48:36like that. So that's in the configure tab. So that's it. And it's like you can you can create a custom GPT literally
613. 1:48:42within minutes. So, you just go to GPTs, the little GPTs thing on the left, and then in the upper right hand corner
614. 1:48:50right here, you say create, and then you're off to the races. All right. All righty. James Cameron doesn't need actors
615. 1:48:59anymore. I know, but neither do any of us. This is a democratizing tool. Listen, just like it used to be, if you wanted
616. 1:49:11to shoot a film, you had to have enough money to go buy canisters of 35 millimeter film and cameras that could
617. 1:49:20hold that film and operators that knew how to work that film. And then you had to have enough money to have that film
618. 1:49:25developed. And then you had enough money to have an editor who had a big giant machine to slice together 35 millimeter
619. 1:49:32film and track the audio and do the and then and then they came out with Final Cut Pro. And they're like, "Oh, that's going to
620. 1:49:40ruin the film industry because now any Yahoo with a [ __ ] Mac can make a movie. What about us?" Same [ __ ] same panic.
621. 1:49:54Storytelling doesn't change. The tools are changing. How you produce [ __ ] changes. It just happens to make it
622. 1:50:03more accessible to literally millions of people. But it's still about telling compelling stories. And if you can tell a
623. 1:50:11compelling story, whether you use film or video or AI, it's a compelling story. Like it's nothing changes. Nothing
624. 1:50:21changes except everything changes. So the existing way of doing things in every job imaginable is about to be transformed.
625. 1:50:33All right. Uh we don't need James Cameron anymore. Exactly, Robbie. Exactly. Kyle, did you figure out the private setting for lovable? Um I
626. 1:50:43didn't. Let me see if they got back to me. I did put a request for someone to [ __ ] help me on the lovable thing questions. Uhuh. [Music]
627. 1:51:00Oh, they're never going to answer everything because now that they're doing this contest. No, they didn't get back to me. Let me go look.
628. 1:51:28project cardboard. Let's see if they fixed the bug. So, I'm going here. I'm going to settings. It's got workspace. I'm going
629. 1:51:37to change it to public. I just changed it. says visibility updated. Your project is now public. Now I'm going to switch out of this page
630. 1:51:47and come back to it. And it still says workspace. I do not how I do not know how to make this setting stick. I don't
631. 1:52:01I've since quit and restarted my machine and Chrome. The only thing I could think is I go settings,
632. 1:52:18public, publish, update.
633. 1:52:34Project updated. Try it now. Brandon, try it now. If it if it's still not there, like I I
634. 1:52:50don't I don't know. Gaffin Biddy, some of us has have lost interested in people. Nothing. I don't know what to say. Brandon, I I um I put
635. 1:53:03a request into the lovable Discord and they didn't get back to me. I'm gonna I think I'm going to DM Tisha.
636. 1:53:12I'll I'll work on that this weekend. I'll see if I can get that fixed this weekend. Um All right. Let's go back to Flo. Okay.
637. 1:53:19Now we have something. What I was hoping to do was have a video we could put up on the the internets.
638. 1:53:36[Music] This is pretty good. Hey, Giants, I'm 6 in of pure doom. Kneel before me, Redwoods, now hand me a ladder. This
639. 1:53:45lens insults my stature. Hey, Giants, I'm 6 in of pure doom. Kneel before me, Redwoods, now hand. That's pretty good.
640. 1:53:53All right, download. And then we'll upres this to 1080p. Okay. So now what I promised was
641. 1:54:07in irregulars. I'm going to upload share cool things. I'm going to upload
642. 1:54:23here is the PDF. of the um deep research
643. 1:54:41guide to VO3.
644. 1:54:59All right. So, don't notify there. So, that's there. And then in AI tool talk, which is behind the mastermind firewall, I'm going to share
645. 1:55:15the project. So, let me go to the project video scene builder. So, can you share a project? Maybe you
646. 1:55:29can't. Hang on. I can rename it. I can delete it. I don't think I can share it.
647. 1:55:43H. But what I'll put in what I'll put in tool talk is I'll put that that thing that we actually put some time into.
648. 1:55:55Oh, wait. You're not seeing any of this. God damn that thing. I find the GPTs I want, but I can't
649. 1:56:10invoke it on my chat. I don't think you can make a GPT if you're just cosmic. If you're just using the mobile app, I don't think you can
650. 1:56:19create a GPT on the mobile app. I think only on desktop. No export option. No, there's nothing. Let me Oh, uh, let me go to a new chat.
651. 1:56:37Um in chat GPT projects, can you share a project with the public?
652. 1:56:58You cannot currently directly share a project. All right. So, I'll just go to um to tool talk and I'll say here is the VO3
653. 1:57:10prompting structured prompting guide. Structured prompting prompting guide. And I'm going to put um with
654. 1:57:26acting instructions and dialogue. Um, this is good
655. 1:57:42to use as a knowledge file for a project in chat GPT to generate prompts
656. 1:57:57without having to use your brain. Yay. All right. And so I'll put the recipe and filmmaker guide in there. This
657. 1:58:12thing's This thing's really cool. Actually, the one thing that I would fix about this is I would fix it so it
658. 1:58:17doesn't add quote marks to dialogue. All right. And then we've got Let's go get our file. Did this download our our dude? No.
659. 1:58:30Download. Okay, there we go. So, let's go look at Oh, he looks great. Okay, so this is the 1080p version
660. 1:58:47of our dude. Hey, giants. I'm 6 in of pure doom. Kneel before me, Redwoods. Now, hand me a ladder. This lens insults my stature.
661. 1:58:59Nice. All right. Beautiful. Beautiful peoples. So let's go to X. X marks the spot. Compose post.
662. 1:59:15Yes. Rather the pilot of the Micro Foot series.
663. 1:59:35Big foots are overrated. Micro foots are the new
664. 1:59:53are all the rage. #Bigfoot
665. 2:00:10Bigfoot AI. There we go. All right. And so we're going to put up [Music]
666. 2:00:24the new version of Whisk should be available for everyone. It's fun. Okay. Yeah, I think I saw that. Is it that's
667. 2:00:31at is it at uh labs labs.google.com? There's flow.
668. 2:00:58Create flow sparkify video whisk. Try it now. New experimental tool that lets you use images as prompts. Visual visual ideas
669. 2:01:09to tell your stories. Enter the tool. That's cool. They added a little guide to the beginning of this. This is pretty
670. 2:01:23slick. All right. So, you put in a picture image, a place image, and whatever that one is. Style. Oh, cool. So, you can either
671. 2:01:37enter Oh, they've really improved the the interface here. This is this is slick. Yeah. So, if you go to labs.googlefxtoolswisk
672. 2:01:45Googlefxtoolswhisk here. I'll put it in the I'll put it in the the YouTube thing
673. 2:01:59authentically. Joe Ultra Plan has V2 credits in Whisk as well. Oh, cool. Nice. All right. Groovy G. Groovy Groovy
674. 2:02:09Groovy. I like it. All right. Did that upload error. No. Why would it upload? Why, Elon? Why? Why? Why can I not upload a [ __ ] file to
675. 2:02:26your stupid [ __ ] service? 261 error. Well, let's try posting it. Post
676. 2:02:402 24 40 50 60. Come on, you can do it.
677. 2:02:57Success. All right, everyone. Go to my thing. Go find Bigfoot. Let's start the microfut trend. [Laughter]
678. 2:03:14All right, everybody. Everybody. Everybody, I'm going to get out of here. Have a fantastic weekend. Um, here's the thing.
679. 2:03:27So, for homework this weekend, um, if you were here for most of the night, like what I experimented with is going from
680. 2:03:39just prompting informally to creating a more structured tool to help you prompt. Like, I would just play around with
681. 2:03:46that. Just play around with the idea of any place where you're sort of halfassing your prompt, get chat GPT to
682. 2:03:55help you do that in a in a more sophisticated way, like whatever that means. Um, discover the power of unplugging
683. 2:04:05your brain and letting it do the hard work and just surfing at this higher level. Um, I I I think there's some real some very deep deep deep
684. 2:04:16power there. And I think in the future as these tools get more sophisticated, they're going to naturally do a lot of
685. 2:04:22the stuff for us. We we just won't even know what's happening on the back end and it'll be doing all this anyway. But
686. 2:04:28if you want to if you want some of the advantages of that now, you have to you have to dig a bit deeper. So that might
687. 2:04:34be something fun to play with this weekend. All right. Um happy Father's Day. Thank you very much. Um hope you
688. 2:04:41have a fantastic weekend. I will be back Monday. Um, next week we've got uh the AI salon um meet and greet. So, if you
689. 2:04:53want to meet people in this community, come to those meetings. Getting to know people in this community is incredibly
690. 2:05:00important. The meeting we had today uh in the mastermind, the mastermind um live session that we have on Fridays now
691. 2:05:07um was a conversation of a different nature. It was it was a much more specific, much more tactical people
692. 2:05:14really helping one another. Like I really like what's starting to already emerge there. We just started it. So
693. 2:05:20super excited about that. Um all right, have a fantastic weekend everyone. Happy Father's Day to everybody and I will see you next week.